Evaluation of HRV Prior to Post-CABG Atrial Fibrillation via a Predictive Analytics Platform

Introduction: Post-operative atrial fibrillation (POAF) affects up to 40 percent of coronary artery bypass graft (CABG) patients and is associated with stroke, prolonged hospitalization, and increased mortality. POAF typically occurs between one and six days after ICU admission, peaking on postoperative day two. Early identification of at-risk patients remains an unmet need. We hypothesized that frequency-domain heart rate variability (HRV) metrics derived from ECG waveforms within 24 hours of ICU admission can be used to distinguish patients who later develop POAF by detecting early autonomic imbalance.


Methods: With IRB approval, we retrospectively analyzed de-identified ECG waveform data from 58 post-CABG patients monitored in a cardiac ICU via a Predictive Analytics Platform (Nihon Kohden Digital Health Solutions, Irvine, CA). HRV metrics were extracted from one-hour ECG segments during the first 24 hours of ICU admission. R-waves were detected using a modified Pan-Tompkins algorithm and verified for sinus rhythm. R-R intervals were resampled using PCHIP interpolation. Frequency-domain metrics including low-frequency (LF percent), high-frequency (HF percent), and LF to HF ratio were computed using the Welch method. Differences between POAF and non-POAF groups were evaluated using Wilcoxon rank-sum tests.


Results: POAF occurred in 22 of 58 patients (38 percent) between days 1 and 6, with a mean onset of 2.2 days. Compared to those without POAF, affected patients had significantly lower LF percent (64.1 vs. 86.8, p = 0.004), higher HF percent (35.9 vs. 13.2, p = 0.004), and a lower LF to HF ratio (1.8 vs. 6.6, p = 0.004). Mean R-R intervals were not significantly different (p = 0.831), supporting autonomic imbalance rather than heart rate as a differentiator.


Conclusions: Frequency-domain HRV metrics obtained early in the ICU course are associated with subsequent POAF, revealing autonomic signatures in advance of arrhythmia onset. These robust, scalable metrics are suitable for real-time implementation and support HRV-based early warning models for post-CABG patients.